AN EXAMINATION OF THE YOUTH VOTER PARTICIPATION RATES WITH INDIVIDUAL LEVEL AND ELECTION SPECIFIC INFORMATION Except where reference is made to the work of others, the work described in this thesis is my own or was done in collaboration with my advisory committee. This thesis does not include proprietary or classified information. _______________________________________ Holly L. Meads Certificate of Approval: ________________________ ________________________ T. Randolph Beard John D. Jackson, Chair Profesor Profesor Political Science Economics ________________________ ________________________ Richard Ault George T. Flowers Associate Professor Dean Economics Graduate School AN EXAMINATION OF THE YOUTH VOTER PARTICIPATION RATES WITH INDIVIDUAL LEVEL AND ELECTION SPECIFIC INFORMATION Holly L. Meads A Thesis Submitted to the Graduate School of Auburn University in Partial Fulfillment of the Requirement for the Degree of Master of Science Auburn, Alabama December 19, 2008 iii AN EXAMINATION OF THE YOUTH VOTER PARTICIPATION RATES WITH INDIVIDUAL AND ELECTION SPECIFIC INFORMATION Holly L. Meads Permission is granted to Auburn University to make copies of this thesis at its discretion, upon request of individuals or institutions and at their expense. The author reserves all publication rights. ________________________ Signature of Author ________________________ Date of Graduation iv VITA Holly Lynn Meads, daughter of Clifford and Patricia Meads, was born on December 2, 1984 in Houston, Texas. She graduated from Armstrong Atlantic State University in May 2007 with a Bachelor of Arts in Economics. She then entered the Master of Science in Economics program at Auburn University in August 2008. v THESIS ABSTRACT AN EXAMINATION OF THE YOUTH VOTER PARTICIPATION RATES WITH INDIVIDUAL LEVEL AND ELECTION SPECIFIC INFORMATION Holly Meads Master of Science, December 19, 2008 (B.A., Armstrong Atlantic State University, 2007) 63 typed pages Directed by John D. Jackson The young voters (age 18 to 24) political participation by voting has declined in the previous Presidential election years and few studies have determined the cause. This study examines the youth participation rates in Presidential elections 2000 and 2004 by using election specific and individual level information. This current study investigates the changes in the voter participation rate between the two years. The results show that the changes in the youth vote are a function of race, marital status, and specific state legislation pertaining to the election. vi ACKNOWLEDGEMENTS The author would like to thank Dr. John Jackson who offered time and help pertaining to the data and econometrics assistance while writing this thesis. The author would like to continue to thank her parents, Clifford and Patsy Meads, and siblings, Caroline, Wilson, and Elizabeth Meads, who gave their support before, during, and after the planning and writing of this thesis. vii Style manual or journal used: American Economic Review. Computer software used: LIMDEP, Microsoft Excel, Mac Excel, Mac Word, and Microsoft Word. viii TABLE OF CONTENTS LIST OF TABLES ............................................................................................................ ix I. INTRODUCTION ...........................................................................................................1 II. LITERATURE REVIEW ...............................................................................................4 III. THE ANALYTICAL FRAMEWORK .......................................................................18 IV. ESTIMATION AND RESULTS ................................................................................30 V. CONCLUSION ............................................................................................................42 REFERENCES .................................................................................................................45 APPENDIX I ....................................................................................................................47 APPENDIX II ...................................................................................................................53 ix LIST OF TABLES Table 1. Voter Participation Between Age Groups ............................................................3 Table 2. Definitions and Descriptive Statistics .................................................................29 Table 3. Binomial Probit Estimation of ES Variables ......................................................33 Table 4. 2SLS: Voter Participation of Ages 18-24 ...........................................................35 Table 5. 2004 Results from Restrictions of 2SLS ............................................................ 38 Table 6. OLS: Voter Participation of Voters Age 18-24 ..................................................53 Table 7. OLS: Youth Voter Participation 2000 & 2004 ...................................................54 1 CHAPTER I INTRODUCTION The voter participation rates of the youth (ages 18 to 24) have been considered bleak in the past thirty years. Although some say that this may portend a lack of political behavior in the future by this generation, the electoral participation of the youth has always been lower than that of the general population. In 1971, the 26 th Amendment was ratified which dropped the voting age from 21 to 18. This new amendment went through the fastest ratification process in U.S. history- only three months. The first Presidential election, in which 18, 19, and 20 year olds could participate in was in 1972 between Nixon and McGovern. This election welcomed over 12 million newly enfranchised voters to the scene (all draft eligible- except females) during a heated election over the future of the Vietnam War. Nearly half of the voters age 18 to 24 voted in that election which was eventually won by Nixon. This phenomenal number of newly enfranchised voters may have been responsible for the discontinuation of the military draft in 1973 by the U.S. government. However, since that time, young voter participation rates have declined. In studying this phenomenon, Angello (1973) stated that in ?a variety of countries at diverse periods in history? young voters have the lowest level of participation in the electorate? (p. 253). Although this observation was made just two years after the 26 th Amendment was ratified to the U.S. Constitution, the phenomenon has been observed in every 2 Presidential election since. In just the past four Presidential elections the 18-24 age bracket voted in smaller percentages than any other age category (e.g. a difference of 16.8% from the second lowest bracket in 1992). The difference in the youth vote compared to the other age brackets can be seen in Table 1 which accounts for the past seven Presidential elections. While the voter participation rates are lower, it is this age group who are more likely to organize and lead political protests. Discussing the lack of voter participation by the youth, Jane Eisner (p. 97, 2004) notes: ??considering the age of these kids, it?s no wonder. They began to walk and talk when Ronald Reagan was President and the era of big government was declared over. They came of age during Bill Clinton?s time, when the Oval Office became a late-night TV joke and faith in the political system took more blows than a prize-fighter on a very bad day.? Eisner also argues that the lack of political participation may be from the extensive state legislation behind voter registration laws and polling times (which are not necessarily hindering just youth voters, but the general population). Another study shows that the youth are less inclined to vote when there is severe party polarization within the Electoral system (Fieldhouse et al., 2007; Cebula and Meads, 2007). Wolfinger and Rosenstone (1980) point to economic and socioeconomic reasons for a lower voter turnout. The purpose of this study is to examine the voting behavior of 18 to 24 year olds in the 2000 and 2004 Presidential elections. The main focus of this study is to examine the effects of voter registration/polling legislation and political party dominance on the young citizens voting behavior. There have been a number of studies on this topic which 3 have explored the effects of economic, socioeconomic and election specific variables on political participation of the general population. This study will apply the theory pertaining to the general population to the younger generation. Table 1: Voter Participation Rates Between Age Groups 18-24 25-44 45-64 65+ 2004 41.9 52.2 66.6 68.9 2000 32.3 49.8 64.1 67.6 1996 32.4 49.2 64.4 67.0 1992 42.8 58.3 70.0 70.1 1988 36.2 54.0 67.9 68.8 1984 40.8 58.4 69.8 67.7 1980 39.9 58.7 69.3 65.1 Source: U.S. Census Bureau, Current Population Survey, Table A-1 This study is divided up into five chapters: Introduction, Literature Review, the Analytical Framework, Estimations and Results, and the Conclusion. The Literature Review examines the historic theories behind voter behavior and empirical studies of voter turnout. The Analytical Framework chapter develops a model based on those theories and studies, defines each individual explanatory variable, and discusses in detail the anticipated theoretical behavior of each variable. In the fourth chapter, the structural equations are presented and estimated using two-stage least squares. The results are then analyzed in terms of the sign and significance of each variable?s coefficient. The concluding chapter summarizes the study?s findings and present ideas for further research in the area. 4 CHAPTER II LITERATURE REVIEW This chapter highlights previous studies which pertain to the young generation?s voting participation. While some studies mainly focus on the total population?s voting participation, the same assumptions can be applied to the younger generation?s voting participation. The theories in the beginning of this section pertain to the electoral participation of the general population and the studies noted in the later part of this section are concentrated on youth-related voting participation. There have been a number of prominent studies on voter participation rates. The most notable of these is the rational voter model by Downs (1957). Other notable studies pertaining to voter turnout of the total voting population were performed by Buchanan and Tullock (1962) and Wolfinger and Rosenstone (1980). The theories they proposed were applied to younger generation voting in studies by Jane Eisner (2004), Angello (1973), Fieldhouse et al. (2007) and Highton and Wolfinger (2001). Anthony Downs (1957) examined voter behavior pertaining to turnout and party choice. After explaining a few of the model?s assumptions (for example: only two parties to choose between, only one party can win, only one vote per citizen, etc.), Downs explains that the decision process to vote or to abstain is a rational process which weighs the costs and benefits of voting. Therefore, the voter will only participate in the election when the expected marginal utility of participation is higher than the expected marginal 5 cost of participation. Downs notes, ?? every rational man decides whether to vote just as he makes all other decisions: if the returns outweigh the costs, he votes; if not, he abstains.? (p. 260) Downs comments that if one?s marginal benefits are equal to the marginal costs, then the voter is indifferent and abstains. An indifferent citizen who votes has made an irrational decision according to Downs. According to Downs, the benefits and costs of voting differ from person to person. The study goes on to identify what one should consider as a benefit of voting. One benefit of voting to an individual citizen is dependent on the political party in which the citizen supports. For example, if one potential voter supports certain policies that have been noted as trivial to the candidate running for the Presidency, that citizen may have an increased benefit of voting against that candidate. On the same note, a citizen may have an increased expected benefit from voting if the candidate supports nearly all of the potential voter?s favored policies (which is uncommon in politics). In another case, the potential voter may have a decreased benefit of voting if he or she is not aware of the policies that are supported or ignored by the candidates. Downs argues that, ?the more information a citizen receives about the policies of each party, the less likely he is to be indifferent.? (p. 264) Downs also comments, ?Only benefits which voters become conscious of by Election Day can influence their voting decisions; otherwise their behavior would be irrational.? (p. 38) While discussing the benefits of voting, Downs also explores democracy and the decision to vote itself. He notes that democracy is dependent on the level of importance that citizens place on their individual votes. Downs observed: 6 ?The importance of his own vote depends upon how important other people think their votes are, which in turn depends on how important he thinks his vote is. He can conclude either that (1) since so many others are going to vote, his ballot is not worth casting or (2) since most others reason this way, they will abstain and therefore he should vote.? (p. 267) Downs then notes the problem in this statement (given the correct assumptions): Democracy is then unable to function rationally. As a solution, it is then apparent (from democracy actually lasting this long) that the citizens may feel a sense of social responsibility to participate in the Electoral process which is ?independent of their own short-run gains and losses.? (p. 267) Citizens may accept this social responsibility and may receive a return (or benefit) from fulfilling this sense of social responsibility. Downs then explores costs of voting based on the assumptions in his model. Since poll taxes are not constitutional, the only direct cost of voting is the transportation to the polls themselves. The indirect cost of voting consists of time at the polls. One citizen?s time may be of more valuable than the others. For example, time costs are inversely related to income. Downs observes, ?If the time must be taken out of working hours, this cost can be quite high, in which case high-income groups again have an advantage. But if the time comes during leisure hours, there is no reason to suppose any such income-correlated disparity exists.? (p. 266) His study claims that the wage lost by voting (a direct cost) from transportation costs is easier to bear by the high-income group rather than the low-income group. Although, recent literature has shown that the opportunity costs of time are higher for the wealthier citizen than that of a lower-income voter. This finding is inconsistent with Downs. 7 The next section of Downs? Economic Theory of Democracy is devoted to explaining voting behavior based on the political parties in which the citizens have a chance to vote for or against. Downs explains this voting behavior as the ?party differential.? (p. 261) The party differential is defined as the expected marginal benefit of voting for the certain party minus the expected marginal cost of voting. Just as before, if the party differential is equal to or less than zero then the voter abstains from voting. Downs then explains, ?If his party differential is not zero, he estimates how close the election will be and discounts his party differential accordingly.? (p. 272) This factor is very important for this current study. If the voter expects the election to be very close between the two parties, then he or she will have an increased benefit from voting and therefore will cast a ballot. In contrast, if the voter expects the election, to favor one party by a landslide, then the benefit is decreased from voting (from his or her vote counting less) and therefore will not cast a ballot in the election. The expected marginal costs and benefits, by definition, may seem trivial on a definition basis. Downs digresses, ??the returns from voting are usually so low that tiny variations in its cost may have tremendous effects on the distribution of political power.? This argument is the catalyst for the policy of Election Day registration, longer polling hours, or even holding elections on national holidays. These cost-reducing options may increase voter participation, according to Downs. Downs? study is the cornerstone of voter behavior. Since his book, An Economic Theory of Democracy, voter behavior has been examined on many different levels, although all the studies base electoral participation on the expected marginal benefit and expected marginal costs of voting. A book entitled The Calculus of Consent (1965) by 8 James Buchanan and Gordon Tullock picks up nearly exactly where Downs finished in An Economic Theory. Buchanan and Tullock (1965) explore voter behavior pertaining to the act of voting itself and the party chosen but incorporate less stringent assumptions than Downs and therefore analyze voter behavior through game theory. This body of work is also well known, as is Downs?, and often the basis for many voter behavior models. Buchanan and Tullock introduce the notion of ?logrolling? into voter behavior. Logrolling is an expression describing a practice of exchanging favors (or in this case, votes) among the general population. In other words it is the previous agreement by two people of reciprocal voting for each other?s legislation. Buchanan and Tullock comment on voting behavior as introduced by Downs (expected benefits versus costs) and then present, ?When any consideration or more inclusive voting rules is made, the incremental costs of negotiating bargains must also be taking into account.? (p. 143) The theory then goes on to introduce game theory into electoral participation, and the subsequent voter behavior model introduced in The Calculus of Consent is very detailed and beyond the scope of this study. This current study closely mirrors that of Raymond Wolfinger and Steven Rosenstone (1980). After noting the impediments of conventional political participation models, Wolfinger and Rosenstone control for certain socio-demographic variables and then study the political participation of each demographic group. Wolfinger and Rosenstone comment on previous studies: ?Nearly all existing research on [voter] turnout has assumed, in both the substantive theory and the statistical models employed, that the effect of a 9 particular variable is the same for all types of people. This is probably not true? the marginal effect on the probability of voting is probably not constant across individuals.? (p. 9) Therefore, Wolfinger and Rosenstone studied different demographics, such as race, age, sex, and education, and identify a model for voter participation. Wolfinger and Rosenstone used probit analysis to estimate the effect of socioeconomic variables on voter turnout. The data consisted of the 1972 Presidential election and the 1974 Senatorial election and was derived from the U.S. Census Bureau and the Statistical Abstract of the United States. Wolfinger and Rosenstone noted (from Downs, 1957) that political participation was a measure of the expected marginal benefits to the expected marginal costs. This consists of many different aspects of socioeconomic, economic, and state suffrage legislative variables. In their model, Wolfinger and Rosenstone were able to control for different demographics of each voter and were able to identify the effects of different variables. For example, with first glance at the voter participation between ethnic backgrounds, more specifically blacks and whites, the statistics indicate that whites were voting more than blacks. This could be indicative of many aspects, but after controlling for education, region, and other socioeconomic variables, the results showed that blacks were voting at a higher rate than whites. Wolfinger and Rosenstone note, ?This has been attributed to blacks? sense of racial identity and to their greater community consciousness.? (p. 91) Another noteworthy result is that of marriage and political participation. Wolfinger and Rosenstone found that there was a positive significant relationship 10 between marriage and voting behavior for every age and educational group. The study observes, ?People with very little autonomous political motivation are most likely to respond to political stimuli from those with whom they have continuing relationships. Moreover, marriage provides a setting for the reinforcement of one?s own beliefs.? (p. 45) This result indicates that the aspect of living with someone stimulates conversation concerning current political issues at hand, and in turn, supports the age old notion that voting is a social behavior. The study continues to note that the results find that couples who are married not only are more likely to vote together, but also vote in favor of the same topics or even abstain from voting together. Wolfinger and Rosenstone continue to note that it was uncommon for one spouse to vote and the other not; usually both voted or both abstained from doing so. In regards to age, Wolfinger and Rosenstone found that the younger population did not vote in the same magnitude that the older generations did. The study provided support for results found in a study by Converse and Niemi (1971), which showed that the younger generation was more apathetic to political participation because of other aspects in their life. They claim that the youth are more concerned with finding a mate and a suitable career than with voting. Wolfinger and Rosenstone then show that an increase in political participation may come about from the youth taking on ?adult roles? later on in life. The study proposes that once the young citizen has found a mate and a career, his or her mobility decreases and their focus is then turned towards political involvement. After looking at the effects of socioeconomic and economic variables, the study also examined the effect of certain pieces of legislation that were enacted in different 11 states. Wolfinger and Rosenstone looked at the effects of the state registration laws enacted during the time period and noted, ?Registration is usually more difficult than voting, often involving more obscure information and a longer journey at a less convenient time, to complete a more complicated procedure.? (p. 61) The study consisted of the laws pertaining to periodic registration for citizens within one state, residency requirements in order to register within a state, closing dates of registration for election days, and national standards for absentee ballots. The three variables in which held a significant impact on voter participation consisted of the state?s registration closing date, hours of operation for registration offices, and the availability for absentee ballots. Wolfinger and Rosenstone then argued that in order to maximize voter participation in any type of election, the state would have to eliminate registration closing dates and extend the hours of operation of registration offices. The three previously discussed studies pertained to the general population of the United States. As stated before, the main focus of this study is the voter behavior of the population age 18 to 24. The following studies take the theories by Downs, Buchanan and Tullock, and Wolfinger and Rosenstone into account and apply them specifically to the younger generation of voters. The initial interest in this study was spurred by the book entitled Taking Back the Vote; Getting American Youth Involved in Our Democracy by Jane Eisner (2004). Eisner notes the tradition in her family which relates voting back to the timeless metaphor that voting is a social activity. Eisner notes that politicians do not speak to the ?under 30? category because they do not vote. She argues that politicians do not want to speak to young citizens about issues they may be interested in because those citizens do not hold a 12 certain voting participation level. But, in a time where causality is questioned daily, are the youth not voting because politicians are not trying to associate with them or are politicians not associating to the youth because the youth do not vote? She continues on with the study to find why the youth are apathetic towards voting. Eisner found that the youth of today are more interested in community service than expressing their right to vote. Eisner observes that colleges (mostly in California) are reporting that 82.6 percent of incoming freshmen have reported frequent or occasional volunteer work in 2001, while in 1986 the percentage was just 66 percent. The increase in interest in community service rather than voting may rise because the youth do not need to be a certain age to volunteer. But, a complication of volunteering non-voters arises. Eisner notes that in order for community service to work effectively, the workers must in turn vote for the politicians who are calling for a change and hold them to their word. Steven Culbertson (2004) of Youth Service America explains, ?I tell the young people I work with that if you volunteer but don?t vote, twenty years from now your kids are going to be cleaning the same dirty rivers and tutoring in the same lousy schools.? (p. 76) Eisner goes on to examine the apathy in youth voting. The study claims that the youth are getting told that hating politics is considered ?cool.? One must also remember the age of the voters in question and then recall the Presidential terms which were important in their formative years. While discussing the lack of voter participation by the youth, Eisner notes (p. 97), ??considering the age of these kids, it?s no wonder. They began to walk and talk when Ronald Reagan was President and the era of big 13 government was declared over. They came of age during Bill Clinton?s time, when the Oval Office became a late-night TV joke and faith in the political system took more blows than a prize-fighter on a very bad day.? Eisner notes the increased apathy of politics may come from increased costs associated with voting for the youth. For instance, there are seven states in which the first time a citizen votes, he or she must vote in person; his or her absentee ballot will not be counted. This would hinder the 18 year olds who are away from home at college who registered to vote where their parents lived. Another increased cost is restrictive voting hours. Some states require that polls must close at 7:00 p.m. or earlier. This is time that the 18 to 24 year olds are likely to be working or in class. How is one to vote when the citizen is busy during those restrictive hours? Eisner also notes that states which have Election Day registration which may facilitate youth voting because it helps those who are too busy and cannot remember to register by the deadline that a majority of states hold. An argument against Election Day registration, though, is that it would increase voter fraud, which would come in the form of deceased citizens voting for certain politicians. No such evidence has surfaced for these few, select states that are allowing Election Day registration. Fieldhouse, et al. (2007) studied the impact of election-specific and individual- level information on the turnout of young voters (age 18 to 24) at the polls in 22 European countries between 1999 and 2002. In the beginning of the study, the authors explain the importance of voter turnout: ?Turnout matters not only because it is an obvious indicator of political participation, but also because it may be regarded as a democratic health check and is the most all-embracing form of political activity.? (p. 797) 14 The study then took notice of the diminishing turnout of the younger generation at the polls. When the youth grow up to continue on with democracy, how will the ?health check? look? Fieldhouse, et al. studied the ease of involvement in politics, socialization (how often one socializes with peers), civic volunteering, and the economic satisfaction of the youth in the country one lives in. Fieldhouse notes that the most important factor to study in the model would be the character of the individual country?s election rather than the character of the individual voters. Although, both are studied in their model, the main focus was on the different attributes of the elections which are determined by the specific country. The independent variables in their study consisted of economic, socioeconomic, and political participation measures and are then divided up into election-specific and individual-level information. The variables from their study, which are also used in this study, are ethnicity, marital status, and the political interest of the voters. The results show that young ethnic minorities are less likely to vote, possibly due to feelings of disenfranchisement associated with misrepresentation in the government. It also shows that youth with an interest in the politics in the election are more likely to vote. Fieldhouse, et al. studied political interest by the closeness of the election, which was measured as the difference in votes between the winning and second-placed parties in the selected elections. The authors suggest that the variable is positive and significant because political interest is considered as expressive of the benefits of voting. Therefore, the larger the expected margin of victory, the higher the voter participation of the youth. This result is somewhat controversial, though, as it is counter to the theory of party differential advanced by Downs (1957). 15 The primary idea surfacing is that the election-specific and individual-level information is country-specific. The study notes that the youth participation is not just dependent upon the individual-level information but on the election-specific information, including the laws and regulations regarding elections in the individual country. The results show that the model should study one country at a time, and not a cross-section of a region in the world. Angello (1973) performed a dynamic analysis of the youth vote. The study consisted of data from the Survey Research Center for the Presidential Election years of 1952, 1960, and 1968. The study grouped ages (by eight years) into six categories and studied the respondent?s feelings of powerlessness within the political spectrum. Angello?s main focus was to find whether the feeling dissipates or congregates with age. The model found that the most pronounced feelings of powerlessness in the political scope were more prevalent among the youth and the elderly of the population. The study found that the feelings of powerlessness in politics flatten around middle age. The results then note that the election?s outcome may be in the hands of the middle-aged voters. This suggestion is implied because the middle-aged voters have less perceived feelings of powerlessness and consequently tend to show up at the polls more often than the young and the elderly. Angello observes that the feelings of powerlessness do not definitely mean a decrease in voting behavior. The study suggests, ?Instead, withdrawal from voting may reflect a definition of other types of political action including extra-legal and illegal protests, as having greater potential impact on the system.? (p. 258) These feelings of powerlessness through the age spectrum (as were reported by the respondents in the 16 survey) were consistent even after adjusting for education and gender. The study then found that, although participation rates among the elderly are the highest, it is that age group that feels the most powerless as opposed to the lack of participation among the youth who feel less powerless. Angello finds that the lack of voting from the youngest generation may be from feelings of disenfranchisement from the government because of the discomfort of their own economic or social situation coupled with the feelings of powerlessness. The study then goes on to note the increase in political participation with age (by voting) as, ?[the young generation] escape not only legal but also social barriers to the use of conventional political power, gain a ?stake in society? and move toward the acquisition of that power.? (p. 256) This may imply the same result that Wolfinger and Rosenstone found that once the young citizen finds a husband or wife, buys a house, and stays in a centralized area of the country the voter behavior increases. This may be considered as the transition to adulthood for the young voter. Highton and Wolfinger (2001) pick up at this notion of ?adult-roles? and their influence of voting behavior of the citizens from the ages of the early twenties to thirties and forties. Highton and Wolfinger (2001) sought to identify the lack of the youth vote and the determinants of the changes which convert an aging non-habitual voter into a citizen who participates by voting in the Presidential elections. The study recognizes that the younger citizen may not vote as much because of other matters at hand in the young citizen?s mind. The authors note, ?Participating in politics? competes against and usually loses to other, more pressing personal concerns for young people.? (p. 203) The study identifies these personal matters as marital status, home ownership, labor force participation, student status, and educational attainment. The data was taken from the 17 1996 Current Population Survey from the U.S. Census Bureau and included 8,049 observations. Highton and Wolfinger separated the sample into age groups (all being under 25 years of age). Each group consisted of two years (for example: 18-19, 20-21, etc). The results show that married young couples are less likely to vote (as in the results of Angello) and citizens in the labor force are more likely to vote. The focus of Highton and Wolfinger?s study was to correctly identify the ?adult- roles? that largely facilitate voting behavior as the citizen gets older. The most significant factor found in the study was residential stability. Young citizens who were living in the same place for one to two years before Election Day had a higher turnout at the polls by six percentage points. This residential stability was even reflected in the 18 year olds who were still living at home with their parents, who were technically residentially stable. One other measure taken into account was home ownership. This variable was positive and significant at the one percent level. This result signifies that when the potential voter holds a stable job and intends to stay in the area for a long time, the more likely that citizen is to register and vote in Presidential elections. If one were to summarize the results from this literature review, one would argue that the youth voting participation depends on the expected marginal benefits and costs (Downs, 1957) associated with voting in the Presidential elections. These marginal benefits and marginal costs can be measured by legislation pertaining to elections in a country (Fieldhouse et al., 2007), feelings of powerlessness within the generation (Angello, 1973), and economic as well as social factors affecting the youth (Highton and Wolfinger, 2001). In the next chapter, the identification of an appropriate model, variables, and hypotheses are stated and examined. 18 CHAPTER III THE ANALYTICAL FRAMEWORK The specification of an appropriate model for this study is crucial. The model must incorporate the crux of existing theories and try not to omit any crucial variables. This section will identify the model to be used in this current study. Each explanatory variable is examined thoroughly with supporting justification from previous studies. A priori expectations concerning the expected sign of the estimated coefficients are developed. The descriptive statistics are presented as well as the data sources for all of the variables. The intent of this study is to examine the determinants of voter participation rates for the 18 to 24 year old age group. Thus an obvious candidate for the dependant variable in the model is voter participation rate (VPR) which is the number of 18 to 24 year olds that voted divided by the total number of 18 to 24 year olds who are U.S. citizens in the population. The fact that this voter participation rate can be viewed as the probability of an 18 to 24 year old voting causes some empirical difficulties since it must lie within the unit interval (truncation, heteroskedasticity, non-normal errors, etc.). To avoid these difficulties, this study employs the logarithmic odds of participation or VPRPOP which is equal to the logarithm of VPR divided by one minus VPR. The rationale is as follows: The probability of participation (VPR) must be between zero and one; but odds of participation [VPRPOP/(1-VPR)] lie between zero and positive infinity; 19 and thus the logarithmic odds (VPRPOP) of participation will lie between positive and negative infinity?thus circumventing many of the difficulties encountered when using a probability measure as a dependent variable. The two principle factors that this model must account for are the expected benefits and costs that face potential voters. From Downs (1957), citizens will vote only when the expected marginal benefits of voting are higher in magnitude than the expected marginal costs. When introducing variables into the system, they must in some way relate to one of these two vital factors. Each variable can then be identified by its effect on the expected marginal benefits and/or costs. Therefore, the classification of variables is made simple by mimicking other studies previously discussed. Following the theory of Fieldhouse, et al. (2007), this model proposes that the logistic voter participation (VPRPOP) rate is a function of individual-level (IL) and election-specific (ES) information. The general model can be represented as: VPRPOP =f(IL, ES) (1) The ES information can be defined the combination of politics, rules and laws, unique to each state, pertaining to Presidential elections. It is hypothesized that ES information can be characterized as affecting the expected costs of an election. Some ES variables will cause a higher cost for the potential voter while other ES information lowers the cost of voting. The ES variables affect the total voting population and are not specific to one age group and will be discussed in detail later in this chapter. The IL information can be viewed as the socio-economic characteristics of the population and can be measured by variables which affect the expected benefits of the potential voter. Such variables on race, educational attainment, marital status, and the 20 unemployment rate are candidates for consideration. These variables affect different aspects of the benefit of youth voting. Some of the IL information can cause increase in the expected benefit of the voting while some may decrease the expected benefit of electoral participation. All of the IL variables are age specific in this study and only relate to the population who are age 18 to 24. The current model specification (equation [1]), based on previous studies, is the starting point for this study. The ES information entails the results presented by Downs (1957), Wolfinger and Rosenstone (1980), and Fieldhouse et al. (2007). The IL information incorporates the results given by Wolfinger and Rosenstone (1980), Angello (1973) and Highton and Wolfinger (2001). Using this prior information lowers the chance that a measurable variable is being omitted from the study. A minority group within a population may feel disenfranchised by the government, by the constant degradation and misrepresentation, and therefore choose not to participate in the Presidential elections based on a perceived lower benefit of voting. There are a number of studies that have noticed this phenomenon (Cebula and Toma, 2006; Copeland and Laband, 2002; Fieldhouse, et al., 2007). The younger generations of the minority may be more sensitive to these feelings and therefore have a smaller probability to vote than the general minority population. Fieldhouse, et al. (2007) found that an ethnic minority in this age group of 18 to 24 is less likely to vote due to feelings of disenfranchisement. The percentage of the 18 to 24 year old population that is African American is termed BLACK in this model. Received theory suggests that as the percentage of the black population increases within this age group the voter participation will decrease due to perceived lower expected benefits. 21 Recognizing IL effects on young voters, Strate, et al. (1989) noted, ?Among young adults, rates of political participation tend to be low. This is likely due to the primacy of such nonpolitical concerns as obtaining an education, finding a mate, and establishing a career.? Other studies notice this ?adult-role? as well (Wolfinger and Rosenstone, 1980; Highton and Wolfinger, 2001; Converse and Niemi, 1971). One of the most frustrating of these adult-roles may be just finding and keeping a job. This frustration may be ?let out? in the poll booths by voters. A number of studies have shown that as the unemployment increases, citizens tend to vote at a higher rate due to restlessness and the frustration (Fieldhouse, et al., 2007; Cebula and Toma, 2006; Connelly and Field, 1944). Angello stated, ?The individual?s political response, or his lack of response, depends both upon the roles he has been allocated in society and on his interpretation of these roles. Studies of alienation have noted that the extent of an individual?s participation in the social system generally is related to his feelings of efficacy or powerlessness.? (p. 252, 1973) An unemployed person may not be satisfied with his or her role in society and may go to the polls with this frustration. With the belief that one is expressing their disgust or disapproval with the labor market, he or she would perceive higher expected benefits of voting. It follows that as the unemployment rate of the 18 to 24 years age group increases (UR), the voter participation will increase. As previously mentioned, Strate, et al. (1989) observed that the quest for a mate may influence this age group more than political participation. If one is already married then one can then devote extra time to politics. Having a partner at home also leads to 22 more conversation about the politics discussed in the elections, creating conflict over or agreeing with the topics, and then could then lead to a higher voter participation rate. Copeland and Laband (2002) found, by using longitudinal data for the total population, that if one were married, then the probability of voting is higher. As previously mentioned, Wolfinger and Rosenstone found the same results, ?People with very little autonomous political motivation are most likely to respond to political stimuli from those with whom they have continuing relationships.? (p. 45, 1980) After discussing the issues with their spouse, the voter may feel an increased benefit from voting. Therefore, this model includes the percentage of the population age 18 to 24 who is married, and theory suggests that as this percentage (MARRIED) increases, the voter turnout of 18 to 24 year olds will increase as well. The level of education that one has attained effects voter participation as well. Therefore, this model includes the variable HSMORE which is the percentage of the 18 to 24 age population who hold a high school degree or more. A variable measuring educational attainment is present in every voter behavior model dating back to Downs (1957). Studies show that citizens who have a high school degree or more are more likely to be informed about the current affairs and political disputes and therefore will be more likely to vote in a Presidential election (Downs, 1957; Wolfinger and Rosenstone, 1980; Cebula and Toma, 2006; Copeland and Laband, 2002; Highton and Wolfinger, 1996). Highton and Wolfinger noted that since 1996, more teenagers attending high schools were showing up at the polls because of ?exhortation in civics classes as well as the efforts of groups like ?Rock the Vote.?? (p. 207, 2001) The introduction of discussions of elections in civic classes in high schools may have shown to the younger 23 population that a benefit can be received from voting. It may be that these civic classes encourage the feeling of social responsibility which Downs (1957) commented on. The citizen who is more aware of the policies introduced within the election may perceive a higher benefit of voting. Theory thus suggests that as the percentage of the 18 to 24 age population with a high school degree or more increases, the voter turnout of the youth will increase as well. Recall that Downs (1957) and Wolfinger and Rosenstone (1980) found that in order to maximize voter participation levels, the state must choose the appropriate set of rules and laws pertaining to registration, voting days, and polling hours carefully. The main focus of this study to investigate the presence of similar effects by using more recent data. In the 2000 and 2004 Presidential elections, a number of states employed legislation that previous studies had shown as hindering electoral participation and therefore implying a higher expected marginal cost to the potential voter. State level voter-registration and polling laws, though created to facilitate or organize voting behavior, may, in fact, hinder voting. Examples of such laws can be found in states which close the election polls relatively early in the evening (around 7:00 p.m.). Eisner notes that ?limits on polling hours are a time-tested way of discouraging working people from casting a ballot, and it?s also a fine strategy to deter young people, who very likely are at a job, in class all day, or only waking up when the rest of us are ready to retire.? (p. 122, 2004). This example of voting limitations by the state is considered an ES factor and therefore increases the expected costs to voting. When a young citizen must rearrange a busy schedule in order to vote in a Presidential election, he or she may be tempted to abstain from the polls just for convenience sake. The change 24 in schedule is considered as an added cost to the voter. In the 2000 and 2004 Presidential election, there were 26 states which closed polls at 7:00 p.m. or earlier. This model includes a binary variable which accounts for states which close polls at 7:00 p.m. or earlier (RSTRVOTE), and it is expected that states in which close polls relatively early may hinder young citizens from voting and therefore would decrease the 18 to 24 year old voter participation rate as well as the participation rate of the older voters. Most states have voter-registration laws to regulate voting on Election Day. An example of such laws involves a voter-registration deadline day prior to Election Day. In some cases, the deadline to register to vote in a November election would be in late September or early October. For young people busy with school, work, and socializing, registering to vote before Election Day (even up to several weeks before) may be the last thing he or she would think of. Many young citizens who do not currently vote on Election Day are not registered to vote. Currently there are only a few states that allow Election Day Registration. These states include Connecticut, Idaho, Maine, Minnesota, New Hampshire, North Dakota, Wisconsin, and Wyoming. There are more states that will implement this policy during the 2008 Presidential election, including Iowa, Montana, and North Carolina. With Election Day Registration, the young voter would not have to take the time out of his or her day in late September to make sure he or she will be able to vote in November. The voter can register and then vote on one day at the same time. Therefore the cost of voting for the potential voter is diminished due to the convenience of registering and voting all in the same day. This model accounts for these states with a binary variable (EDREG) and suggests that states which allow Election Day Registration will see an increase in the young voter participation. 25 A study by Cebula and Meads (2008) showed that voter participation rate for the total population was a negative function of the dominance level of a political party (either Republican or Democratic) in the Upper House State Legislature. Fieldhouse, et al. (2007), also noted that the ?closeness of the major parties in the polls, party polarization, and the electoral system? were contributing factors affecting the turnout at the polls (p. 798). The study also argues that ?young people may be more sensitive to such factors if, unlike older voters, they have yet to acquire the habit of voting/not-voting.? (Fieldhouse, et al., p. 799, 2007) Downs (1957) also had a notion of party differential. This measure was the difference between the voter?s expected benefit for voting for a particular party and the expected costs of voting. If the party differential was equal to or less than zero, the voter would abstain from voting. With the party differential theory in mind, this current model accounts for political dominance within a state by using the percentage of the Upper House State Legislature which is dominated by either the Republican or Democratic Parties (DOM0003). This percentage is calculated by taking the absolute value (to account for any dominance by either party) of the difference between Republicans and Democrats in the Upper House State Legislature and dividing the difference by the total number of seats. The party-composition of the state legislature is a proxy for the party-composition of the population within a state. As this dominance by any particular party increases, the young voters (both Republican and Democrat) may take on an apathetic attitude towards participating by voting. For example, if Republicans dominate a state, the young Republican citizen may be hindered from voting because the Electoral College is quite likely to be committed to the favored politician of the voter, anyway. Meanwhile, the 26 young Democrat voter will also be hindered to vote because the Electoral College is under the control of the opposing party. This argument therefore suggests that when a certain party dominates the state in a two-party system, the expected cost of voting increases due to feelings of apathy (or ?why bother??). If the citizen feels as though his or her vote will not count as much as in a close race then the voter will be hindered from voting. Therefore, theory suggests that as this dominance percentage increases, the voter turnout of 18 to 24 year olds will decrease. From the explanatory variables introduced above, the model can be specified more concretely. In this study, the IL variables are specified as BLACK, UR, MARRIED, and HSMORE while the ES variables are EDREG, DOM0003, and RSTRVOTE. Therefore, from equation (1) from above, the young voter general model can be made more explicit as: VPRPOP = f (BLACK, UR, MARRIED, HSMORE, EDREG, DOM0003, RSTRVOTE) (2) From the hypotheses above, this study proposes that: f BLACK <0; f UR >0; f MARRIED >0; f HSMORE >0; f EDREG >0; f DOM0003 <0; and f RSTRVOTE <0. From equation (2), the structural equation to be estimated is given by: VPRPOP jt = ? 0 + ? 1 BLACK jt + ? 2 UR jt + ? 3 MARRIED jt + ? 4 HSMORE jt + ? 5 EDREG jt + ? 6 DOM0003 jt + ? 7 RSTRVOTE jt + ? jt (3) where the definitions and descriptive statistics on each explanatory variable is in Table 2. The data available to account for these explanatory variables includes 48 states for the Presidential Election year 2000 and 45 states for the Presidential Election year 2004. The state of Nebraska was eliminated from both regressions due to its unicameral State 27 Legislature (therefore, DOM0003 cannot be computed). Additional states were omitted from the different regressions for lack of data available for certain variables. The state of Connecticut was omitted from the 2000 data set because of lack of unemployment data for this age group, and the states Idaho, Montana, North Dakota, and Vermont were deleted from the 2004 data set for the lack of the BLACK variable for this age group. The source for the explanatory variables VPRPOP, BLACK, MARRIED, and HSMORE, in the 2000 Presidential Election regression, came from the U.S. Census Bureau (Tables 4b, SF1 Table P12B, SF3 Table PCT7, SF3 Table PCT25). The variable UR was obtained through the Bureau of Labor Statistics (Table 12) and the variable DOM0003 was obtained through the 2000 Statistical Abstract of the U.S. (Table 401) for the 2000 Presidential Election regression. For the 2004 Presidential Election regression, the variables VPRPOP, BLACK, MARRIED, and HSMORE were given by the Census Bureau by the Tables 4b, B01001B, B12002, and B15001. The UR variable was acquired from the Bureau of Labor Statistics (Table 14) and the variable DOM0003 was obtained from the 2004-2005 Statistical Abstract of the U.S. (Table 400) for the 2004 Presidential Election regression. The variables EDREG and RSTRVOTE were first found in Eisner?s book Taking Back the Vote: Getting American Youth Involved in Our Democracy, but then were amended by further research into the individual states Board of Elections. In this chapter, the theoretical rationale for including each variable in the model was discussed. This chapter assigned BLACK, UR, MARRIED, and HSMORE into the IL category (which affect expected marginal benefits) and EDREG, RSTRVOTE, and DOM0003 into the ES category (which affect expected marginal costs) of the model. 28 The conclusion of this chapter involved the specification of the major structural equation of the model. In the next chapter, additional structural equations are presented (including additional variables introduced within the next section) and two-staged least squares estimates are obtained for the system. The results are presented with the economic rationale behind the estimated signs and significance of the coefficients is considered. 29 Table 2: Definitions and Descriptive Statistics Variable Definition Mean St. Dev. VPRPOP jt the logit of the voter participation rates of the population age 18 to 24 in state j during Presidential Election year t. -0.335 0.391 ? 0 , ? 0 , ? 0 constant terms BLACK jt the percentage of the population age 18 to 24 who are African American in state j during Presidential election year t. 12.055 11.008 UR jt the unemployment rate of the population age 18 to 24 in state j during Presidential election year t. 10.334 2.519 MARRIED jt the percentage of the population age 18 to 24 who are married and living together in state j during Presidential election year t. 13.861 4.799 HSMORE jt the percentage of the population age 18 to 24 who hold a high school degree or more living in state j during Presidential election year t. 79.379 4.832 EDREG jt a binary variable indicating whether state j allows Election Day Registration during Presidential election year t. 0.140 0.347 DOM0003 jt the percentage of the Upper House State Legislature of state j which is dominated by either the Republican or Democratic Party (during Presidential election year t for 2000 or t-1 for 2004) which is calculated by taking the absolute value of the difference of Republicans and Democrats in the Upper House and then dividing the difference by the total number of seats. 25.050 19.647 RSTRVOTE jt a binary variable accounting for state j which closes polls at or before 7:00 p.m. on Election day during the Presidential election year t. 0.548 0.500 REDST jt a binary variable representing whether state j was determined as a red or blue state in the given Presidential election t. 0.602 0.492 LOGGSP jt the logarithm of the state total real gross domestic product for state j during Presidential election year t. 11.758 1.041 D jt a binary variable representing the Presidential election year 2004. 0.516 0.502 ? jt , ? jt , ? jt stochastic error terms 30 CHAPTER IV ESTIMATION AND RESULTS In this chapter, the structural equations of a more complete model and the results are presented. In the beginning of the chapter, problems encountered in the preliminary regressions are addressed and two new variables as well as two additional structural equations are introduced to the system. Then, the new structural system is estimated by two-staged least squares. Following the estimation, the results are presented, and the reasoning behind some of the structural changes between the 2000 and 2004 Presidential election years is discussed. After some preliminary regressions, the anomalous results suggested that two of the three election specific variables may be endogenous, which results in causality problem. That is, voter participation may influence whether a state has restrictive voting hours or Election Day registration as well as the converse, which was originally specified. To control for this, the brute force two-staged least squares methodology is applied to the implied system. In the first stage, the endogenous variables will be regressed upon several, but not all, of the dependent variables in equation (3). Additional variables were added in the first stage to the system of equations for a more precise estimation. The estimated values of these endogenous variables are then placed into equation (3) and regressed with the remaining exogenous variables. 31 In order to predict EDHAT, and additional variable was needed to explain Election Day registration more precisely. The variable added to this equation was the logarithmic form of the gross state product, which was denoted as LOGGSP. After placing the Election Day registration states on the map, it was noticed that these states where not states in which the GSP is historically very high (Connecticut, Idaho, Maine, Minnesota, New Hampshire, North Dakota, Wisconsin, and Wyoming). Therefore, the LOGGSP variable was expected to be inversely correlated to states with Election Day registration. From the regression results, the predicted values of EDHAT could then be calculated. An additional variable also needed to be added to the equation for predicting RSTRHAT. These states were then also placed on a map and noticed that they were the states which were historically red states in Presidential elections. A binary variable accounting for whether the state was determined as a red state or not during the 2000 and 2004 Presidential elections was added to the system and denoted as REDST. This variable was expected to be directly correlated to states which have restrictive polling hours. Therefore the structural equations for the model are as follows: VPRPOP jt = ? 0 + ? 1 BLACK jt + ? 2 UR jt + ? 3 MARRIED jt + ? 4 HSMORE jt + ? 5 EDREG jt + ? 6 DOM0003 jt + ? 7 RSTRVOTE jt + ? jt (3) RSTRVOTE jt = ? 0 + ? 1 UR jt + ? 2 HSMORE jt +? 3 EDREG jt + ? 4 DOM0003 jt + ? 5 REDST jt + ? jt (4) EDREG jt = ? 0 + ? 1 UR jt + ? 2 HSMORE jt + ? 3 RSTRVOTE jt + ? 4 DOM0003 jt + ? 5 LOGGSP jt + ? jt (5) 32 where the variables VPRPOP, BLACK, UR, MARRIED, HSMORE, RSTRVOTE, EDREG, and DOM0003, REDST, and LOGGSP are defined in Table 2. The variable REDST was derived from the CNN map of red states and blue states for the election results in 2000 and 2004. The variable GSP was taken from the Bureau of Economic Analysis website and then put into logarithm form. The descriptive statistics for these variables are in Table 2 which is located in the previous section. There were a number of additional variables including more election specific variables to account for EDHAT, but none showed any significance. The results are shown in Table 3. The main question surfacing from the structural equations (3), (4), and (5), is if VPRPOP, EDREG, and RSTRVOTE are jointly determined. This approach was initially taken, but the results did not conform theoretical expectations. These results are presented in Tables 6 and 7 in the Appendix II. Alternatively, RSTRVOTE and EDREG can be viewed as being determined independently of each other but jointly with VPRPOP. From the 2000 and 2004 data, there was not one state which allowed Election Day registration that limited polling hours to 7:00 p.m. This shows that these two legislations could be substitutes for each other. Therefore, with this model, EDREG is exogenous to the RSTRVOTE decision and the converse is true for RSTRVOTE. This approach leads to the following results. The results of the estimation of RSTRHAT and EDHAT reduced forms show a hit-rate of 86.02% and 88.17%, respectively. The Chi-squared is significant in both models; showing 45.29 in the estimation of RSTRHAT and 24.61 in the estimation of EDHAT when the critical value is 1.15 at the 95% confidence interval. In the regression 33 accounting for RSTRHAT, the additional variable REDST is shown to be negative and significant beyond the five percent level. The variable LOGGSP was insignificant in the regression estimating EDHAT. Table 3: Binomial Probit Estimation of ES Variables Estimation of RSTRHAT Estimation of EDHAT Coefficient T-ratio Coefficient T-ratio Constant 17.596* 4.29 -0.420 -0.08 UR -0.228* -2.88 -0.237* -2.47 HSMORE -0.211* -4.22 0.099** 1.79 EDREG -0.431 -0.75 RSTRVOTE -0.584 -0.98 DOM0003 0.010 1.08 -0.016 -1.51 REDST 0.647** 1.70 LOGGSP -0.512* -2.20 Chi squared 45.29 24.61 Efron R-sq. 0.49 0.26 Hit-Rate 86.02% 88.17% No. of Obs. 93 93 * denotes significance at 0.05 level ** denotes significance at 0.10 level The results in Table 3 show three variables significant in both estimations of RSTRVOTE and EDREG. The coefficients of the variables REDST and LOGGSP are showing expected signs and are both significant at the ten percent level. The null hypothesis of no effect can be rejected. The results in Table 3 are interesting enough, but the main focus of this study is to explain voting behavior of the young population. As stated earlier, the process of estimation is brute force two-stage least squares. The predicted values of EDHAT and RSTRHAT are now placed on the right side of equation (3). VPRPOP will be regressed upon these two endogenous variables and the remaining five exogenous variables. The new structural equation is now: 34 VPRPOP jt = ? 0 + ? 1 BLACK jt + ? 2 UR jt + ? 3 MARRIED jt + ? 4 HSMORE jt + ? 5 EDHAT jt + ? 6 DOM0003 jt + ? 7 RSTRHAT jt + ? jt (6) For this regression, because of including two consecutive Presidential election years, the Chow Test was performed to see if the data could be pooled. The test proved significant with an F-statistic of 2.378 and the critical value at the 95% confidence level was 2.061. Then, a dummy variable (D), which equaled zero for the year 2000 and equaled one for the year 2004, was multiplied by the independent variables in equation (6). The variables multiplied by D are now interaction terms and are denoted with a D on the end of the original notation. While estimating, the interaction terms will measure the changes which happened in the right hand side variables between the 2000 and 2004 Presidential elections. While including the interaction terms, the variables in which were not multiplied by D will be the coefficients of the variables corresponding to the 2000 Presidential election. The interaction variables will allow for the estimation of the 2004 coefficients and t-ratios from the corresponding 2000 variable coefficients. These interaction terms should have been included in equations (4) and (5) but the states that held Election Day registration and restrictive voting hours did not change between 2000 and 2004. Therefore the changes between the two years could not be computed. For the second stage, the variable D, original independent variables, and the new interaction terms were then included in the OLS regression. The results from this regression are in Table 4. These results show four variables significant at the five percent level and one variable significant at the ten percent level. The R-squared is 0.47 and the F-statistic is 4.57 with significance at the one percent level. A Breush-Pagan test was conducted for 35 heteroskedasticity and resulted in a LM statistic of 32.65 which was significant at the one percent level, which shows evidence of heteroskedasticity and therefore, the White robust standard errors correction was used. Ramsey?s RESET test was also performed and resulted in an insignificant F-statistic of 1.40 (when the critical F-statistic level for 95% confidence is 2.74) which shows that the model has been correctly identified. The Chi- square statistic was 64.38 (before the White standard error correction) and is significant at the one percent level (the critical value being 7.96). Table 4: 2SLS: Voter Participation of Ages 18-24 Coefficient T-ratio Constant -1.735 -0.96 UR -0.018 -0.56 BLACK 0.012* 2.12 MARRIED 0.019 1.47 HSMORE 0.015 0.68 RSTRHAT -0.320* -2.02 EDHAT 0.116 0.41 DOM0003 -0.002 -0.71 D 1.496 0.61 URD 0.013 0.34 BLACKD -0.017* -2.56 MARRIEDD -0.030* -2.12 HSMORED -0.008 -0.27 RSTRHATD 0.247 1.23 EDHATD 0.021 0.07 DOM0003D -0.006** -1.69 R-squared 0.47 F-statistic 4.57 No. of Obs. 93 * denotes significance at 0.05 level ** denotes significance at 0.10 level From the results in Table 4, the individual coefficients for each year can be computed and the structural changes are denoted as well. The coefficients of the first set of seven independent variables are the coefficients of the period in which the D dummy 36 variable equals zero (which is the 2000 Presidential election). The coefficients of the interaction variables (those with ?D? on the end; for example: HSMORED or BLACKD) measure the structural changes in the corresponding coefficients between 2000 and 2004. The coefficients for the 2004 Presidential election can be computed from these structural changes between 2000 and 2004 and the coefficients of the first set (2000). The computed coefficients and t-ratios are in Table 5. The first block of results show that during the 2000 Presidential election, there were two significant variables (BLACK and RSTRVOTE) at the five percent level. There were five variables (UR, MARRIED, HSMORE, EDHAT, and DOM0003) which were shown to be insignificant, i.e. the null hypothesis that the true coefficient equals zero cannot be rejected. The variable BLACK is positive and significant at the five percent level. This is the opposite of the expected sign stated above. This positive sign for the young black population in 2000 may be a signal for the future black voter participation rates. It may show that the feelings of disenfranchisement as shown in earlier studies may have caused more frustration for this age group and may be a driving force to the polls. Angello provides a description of frustration at the polls by, ?The individual?s political response, or his lack of response, depends both upon the roles he has been allocated in society and on his interpretation of these roles.? (p. 252, 1973) Expressing one?s frustration of the past oppression within the system may have drawn more young African Americans to the polls in the year 2000. This movement may be a call for change in representation in the government and may show a future of more African Americans voting. This finding is in a way apparent from the result that Wolfinger and Rosenstone found in a 1980 study. 37 They found that while the white population may have been voting in higher numbers, the black population was voting at a slightly higher percentage. (p. 90) The African Americans who were voting in the 1970?s at a ?slightly higher rate? than the white population are now the parents of the young black population who are studied in this paper. The influence of parents, who grew up during the Civil Rights Movement, may have drawn more young people to the polls. This positive sign of the BLACK 2000 coefficient also may show an increase in interest in the Presidential election after a Democrat presence in the White House for the previous eight years. Because the black population historically votes Democratic, these two Presidential terms could have encouraged the young black population for more political participation. In a study by Griffin and Keane (2006), survey results showed, ?32% of all African Americans identify themselves liberal, 40.8% identify themselves as moderates, and 27.2% identify themselves as conservatives.? (p. 1003) This five percent difference between liberals and conservatives may not be large but it is safe to assume that nearly half of the moderates could be voting liberal in elections. This creates a larger difference and shows that the black population may have been more influenced to vote after eight years of the Clinton administration. Because of the positive significance of the young black voter behavior, the null hypothesis of no effect can be rejected. The variable RSTRHAT is also negative and significant beyond the five percent level which supports the previously stated hypothesis. The young population may find it difficult to vote during the allotted time intervals. This specific variable accounted for those polls which close at 7:00 p.m. or earlier. During the specific time apportionment, the young population are going to school and working (and also waking up later than the 38 general population). This time restriction simply does not fit into the busy schedule of a young person. This negative sign follows the advice of Downs (1957) who predicted that the general population (in which the youth is involved) is less likely to vote when there are higher costs involved. Wolfinger and Rosenstone (1980) also found that additional legislation pertaining to voting hours (and registration) hinder potential voters from participating in the Presidential elections. Because of the strong significance, the null hypothesis of no effect can be rejected. From the results in Table 4 and placing restrictions on the variables and the corresponding interaction terms, the coefficients and t-ratios can be computed for the voter participation of the 2004 Presidential election. The 2004 coefficients can be estimated by adding the coefficients of the 2000 variable and the corresponding interaction term in Table 4. The t-ratios can be found by placing a restriction on the 2000 variable and the interaction term; the restriction being that adding the two corresponding coefficients equals zero (ex. UR + URD = 0). Once these restrictions are in place, the t- ratios are then computed by taking the square root of the F-statistic relating to the restrictions. These results are shown in Table 5. Table 5: 2004 Results from Restrictions of 2SLS Coefficient T-ratio UR -0.005 -0.20 BLACK -0.005 -1.11 MARRIED -0.012 -1.15 HSMORE 0.007 0.37 RSTRHAT -0.073 -0.20 EDHAT 0.137 0.57 DOM0003 -0.007 -2.51 39 From the results in Table 4, several of the changes between 2000 and 2004 are significant. The change in BLACK is negative and significant beyond the five percent level. This change is noteworthy because it causes a significant coefficient in the 2000 Presidential election to become insignificant in the 2004 election. The coefficient for the 2004 election is computed as -0.005 and the t-ratio is -1.11. This negative change from a strong positive turnout could come from disenfranchisement during the Republican administration. As stated previously from another study, only 27.2% of blacks considered themselves as conservative. A conservative President in the White House after a Democrat who would serve eight years may have lessened the benefits of voting for the liberal or left-leaning moderate young black population. As a result, in the 2004 Presidential Election between the Republican incumbent and John Kerry, the black variable slope is flatter and less significant in 2004. The other significant structure change is that of MARRIED. This change is negative and significant beyond the five percent level. The 2004 coefficient equals negative 0.012 and a t-ratio of -1.15. Between 2000 and 2004, the coefficient for the married population went from positive to negative. This significant negative change may indicate something about the issues of politics and married couples. The 2000 coefficient, though insignificant, was positive. The 2000 election was the mark of the end of the second term of President Clinton. Therefore, the candidates in the arena were new candidates, none of which were incumbents. Being new faces, the conversation of the new candidates between the married couple (or any groups of people) who lived together would be more ?entertaining? than that of an incumbent and only new candidate. The two candidate?s policies regarding the economy, health care, education, and defense 40 could create more discussion and/or conflict than an incumbent and a new candidate would. Wolfinger and Rosenstone?s (1980) results are congruent with the positive sign, ?If someone has a weak inclination to vote, the presence of another family member who has the same tendency in the same direction will raise the probability that both will vote.? (p. 45) The 2004 Presidential election consisted of candidates who one was an incumbent and the other was a new contender. The conversation between the married couple may become stagnant when one person knows the other?s opinions about the incumbent and that person may even become apathetic to voting if the other?s view is dominating their own. Hence, the negative sign in the difference of the MARRIED variable and the negative, insignificant coefficient in 2004. The difference in DOM0003D is negative and significant at the ten percent level, and causes a larger (in absolute standards), negative coefficient in 2004. This variable, which was shown to be insignificant in 2000, has a steeper slope in 2004 and is now shown to be significant. The 2004 coefficient equals negative 0.007 and the t-ratio is equal to negative 2.514. From the results in Table 5, this variable is the only variable that significantly affected the youth voter participation rates in 2004. This negative change can be due to the election being one of an incumbent Republican President who may have disenfranchised young voters with the tumultuous years during Operation Iraqi Freedom. This measure of party differential creates a higher cost of voting for the young voters. The young citizens may have lost their unstable faith in the political system and are therefore discouraged from participating in the 2004 election in states with a significant party dominance. 41 This chapter explained the dual causality problem from the EDREG and RSTRVOTE variables and the voter participation rate. It also highlighted the results from the two-staged least squares method of estimation with three endogenous variables. The results were then discussed in detail pertaining to the first set of results as well as the second set of results which were the differences within the two Presidential election years. The next chapter, the Conclusion, provides the culmination of the results of the study as well as suggestions for further research within the field of the youth voter participation. 42 CHAPTER V CONCLUSION As the 26 th Amendment was ratified into law on June 30, 1971, Jennings Randolph stated, ?I believe that our young people possess a great social conscience, are perplexed by the injustices which exist in the world, and are anxious to rectify those ills.? As this study suggests, Randolph is correct as long as the youth hold on to their unstable faith in the political system. This particular study used state-level data from 2000 and 2004 in order to explain the behavior of young voters age 18 to 24 in Presidential elections. The model examined in this study investigated the effect of election-specific and individual-level information, which encompassed socioeconomic and economic variables, on the youth voter participation rates. The study found that the youth voter participation rates in 2000 and 2004 were largely affected by the costs and benefits, namely the election specific and individual level information. After estimating a three endogenous variable system in a two staged least squares methodology, the evidence found from this study shows that the youth voter turnout of 2000 was a decreasing function of restrictive polling hours in state. Youth turnout in 2000 was also an increasing function of the percentage of the young population who are African American. This study was also able to determine the significant changes in the youth participation rates between 2000 and 2004. These changes consisted within the percentage of the young population who are African 43 American, the percentage of the young population who are married, and the political dominance within a state. The advantage in the study was the diverse elections that took place in 2000 and 2004. In 2000, new candidates were in the political arena after an eight year period of holding a Democrat President. The 2004 Presidential election was that of a Republican incumbent and a new Democrat. From these differences, this study was able to find the changes in the measureable variables consisting of individual level and election specific variables. This study took the voter behavior model of the general population and applied it to the youth voters aged 18 to 24. The only significant change that this study found, in respect to sign change between the ages, was that of the percentage of the population who are African American. This study found that in 2000 the young African Americans were likely to vote, even though past studies of the general population have found that the total percentage of the black population are less likely to vote compared to their race counterparts. This study suggests that this may be a sign of the future, or just a sign of elections after a Democratic President after two terms. What is noteworthy though, is that the remaining signs of the coefficients are the same for youth as the general population, based on previous studies. This may suggest that the youth operate under the same conditions of rational voting as the general population. There are many ways in which this study can be improved. The youth voter participation rate actually increased between 2000 and 2004. The results in this study show that in 2004, the only thing affecting the youth voter behavior that was included in the model was the political polarization within a state. This result showed a negative 44 impact to the youth vote. This shows that an important factor which cannot be measured would have encouraged young voters to show up at the polls. This factor may have been the rumor of a draft re-instatement for Operation Iraqi Freedom on the internet. A measurable variable to account for this would be internet access in the home for teenagers and citizens in their early twenties. Another research opportunity which comes from these results could consist of comparing the voter participation rate with elections of which both candidates are new to the Presidential election or if one is an incumbent. As stated earlier, the young population who are married may have voted less in 2004 because of the incumbent President and the same policies being argued. The results may be different because of the incumbent President?s policies concerning the welfare of the county in the year 2004 specifically. As one can notice, an issue which is a hot topic for the young in the 2008 Presidential election is the future of Operation Iraqi Freedom. This similar topic of the future of an unpopular war was shown to be important during the first election opened to 18, 19 and 20 year olds, 1972. A way in which to modify this topic in the future is to study the effects of different socioeconomic variables on youth voting after eight years of a Democrat or Republican in the White House. As stated before, having Bill Clinton in the White House for the eight years previous to the 2000 election may have influenced the young, black population and there may be a difference in the 2008 election after eight years of George Bush. The difference of a Democrat or Republican in the White House for a little less than a decade may influence other races to vote as well as citizens with different incomes and educational attainment. 45 REFERENCES Angello, T.J. ?Aging and the Sense of Political Powerlessness.? The Public Opinion Quarterly, Summer 1973, 37, 251-259. Buchanan, J.M., and Tullock, G. The Calculus of Consent. Ann Arbor, MI: University of Michigan Press, 1962. Cebula, R. and Meads, H. ?The Electoral College System, Political Party Dominance, and Voter Turnout, With Evidence from the 2004 Presidential Election.? Atlantic Economic Journal, March 2008, 36, 53-64. Cebula, R. and Toma, M. ?Determinants of Geographical Differentials in the Voter Participation Rate.? Atlantic Economic Journal, 2006, 34, 33-40. Connelly, G.M., and Field, H.H. ?The Non-Voter- Who He Is, What He Thinks.? The Public Opinion Quarterly, Summer 1944, 8, 175-187. Converse, P.E. and Niemi, R. ?Non-voting Among Young Adults in the United States.? in Political Parties and Political Behavior. ed. W.J. Crotty, Boston, MA: Allyn and Bacon, 1971. Copeland, C. and Laband, D.N. ?Expressiveness and Voting.? Public Choice, 2002, 110, 351-363. Downs, A. An Economic Theory of Democracy. New York, NY: Harper and Row, 1957. Eisner, J. Taking Back the Vote: Getting American Youth Involved in Our Democracy. Boston, MA: Beacon Press, 2004. Fieldhouse, E., Tranmer, M., and Russell, A. ?Something About Young People or Something About Elections? Electoral Participation of Young People in Europe: Evidence From a Multilevel Analysis of the European Social Survey.? European Journal of Political Research, 2007, 46, 797-822. Griffin, J.D. and Keane, M. ?Descriptive Representation and the Composition of African American Turnout.? American Journal of Political Science, 2006, 50, 998-1012. Highton, B. and Wolfinger, R. E. ?The First Seven Years of the Political Life Cycle.? American Journal of Political Science, January 2001, 45, 202-209. 46 Strate, J.M., Parrish, C.J., Elder, C.D., and Ford, C. ?Life Span Development and Voting Participation.? American Political Science Review, 1989, 83, 443-464. 47 APPENDIX I 2000 Data VPRPOP UR BLACK HSMORE MARRIED EDREG DOM0003 AL -0.398 12.5 31.45 72.25 18.06 0 25.71 AK -0.183 14 4.71 76.93 18.19 0 40.00 AZ -1.066 7.26 3.53 69.25 16.60 0 0.00 AR -0.560 10.96 19.57 75.43 22.27 0 54.29 CA -0.523 10.54 6.68 70.74 11.94 0 32.50 CO -0.832 7.03 4.12 75.07 15.49 0 2.86 DE -0.448 10.94 22.04 77.62 9.37 0 23.81 FL -0.603 9.16 19.96 71.74 12.69 0 25.00 GA -0.715 8.31 31.98 70.04 15.03 0 17.24 HI -1.427 11.9 3.27 85.84 11.84 0 77.78 ID -0.584 9.3 0.68 77.31 23.22 1 82.86 IL -0.392 9.94 16.61 75.99 10.96 0 8.47 IN -0.730 8.19 9.38 76.49 14.87 0 28.00 IA -0.124 6.9 2.89 81.45 13.51 0 20.00 KS -0.587 8.55 6.80 78.31 17.52 0 50.00 KY -0.660 9.91 8.97 74.93 20.34 0 5.26 LA -0.252 13.55 37.08 72.31 14.10 0 33.33 ME 0.037 8.33 1.01 78.90 11.09 1 0.00 MD -0.462 9.43 31.59 79.63 8.51 0 40.43 MA -0.390 6.57 6.74 82.21 5.55 0 70.00 MI -0.510 7.89 15.59 76.46 10.58 0 21.05 MN -0.148 6.22 4.48 79.29 10.88 1 19.70 MS -0.266 14.1 42.43 71.33 16.03 0 30.77 MO -0.815 8.14 12.64 76.50 15.49 0 2.94 MT -0.341 10.84 0.59 78.56 14.86 0 24.00 NE -0.784 7.84 7.38 66.70 17.90 0 14.29 NH -0.657 6.8 1.22 77.77 8.06 1 8.33 NJ -0.593 9.59 16.84 76.33 6.96 0 20.00 NM -1.149 12.5 2.35 70.50 16.44 0 14.29 NY -0.667 10.45 17.72 76.05 7.19 0 18.03 NC -0.817 9.78 23.74 74.20 16.25 0 40.00 ND 0.348 6.56 1.30 84.41 12.43 1 30.61 OH -0.495 8.58 12.66 76.81 12.63 0 24.24 OK -0.575 6.38 9.42 74.83 21.73 0 25.00 48 VPRPOP UR BLACK HSMORE MARRIED EDREG DOM0003 OR -0.361 9.93 1.94 74.21 15.83 0 6.67 PA -0.763 9.86 11.59 79.76 8.40 0 16.00 RI -0.379 11.25 6.22 81.28 6.01 0 76.00 SC -0.488 10.88 32.83 74.25 13.96 0 4.35 SD -1.170 5.26 0.96 78.21 14.66 0 37.14 TN -1.146 8.99 19.50 75.13 19.09 0 9.09 TX -0.751 10.36 12.28 68.65 18.48 0 3.23 UT -0.412 5.84 0.84 80.31 26.09 0 37.93 VT -0.858 6 0.95 83.02 8.16 0 6.67 VA -0.292 5.85 22.07 79.37 12.67 0 10.00 WA -0.407 12.38 3.87 75.28 14.62 0 2.04 WV -0.735 11.27 4.23 78.20 18.34 0 64.71 WI 0.032 7.09 6.73 78.86 10.16 1 9.09 WY -0.352 9.09 1.32 79.03 18.85 1 33.33 RSTRVOTE REDST LOGGSP AL 1 1 4.41 AK 0 1 4.63 AZ 1 1 4.49 AR 1 1 4.40 CA 0 0 4.58 CO 1 1 4.60 DE 0 0 4.72 FL 1 1 4.47 GA 1 1 4.55 HI 1 0 4.52 ID 0 1 4.43 IL 1 0 4.57 IN 1 1 4.50 IA 0 0 4.49 KS 1 1 4.49 KY 1 1 4.44 LA 0 1 4.47 ME 0 0 4.44 MD 0 0 4.53 MA 0 0 4.64 MI 0 0 4.53 MN 0 0 4.57 MS 1 1 4.35 MO 1 1 4.50 MT 0 1 4.37 49 RSTRVOTE REDST LOGGSP NE 1 1 4.56 NH 1 1 4.55 NJ 0 0 4.61 NM 1 0 4.44 NY 0 0 4.61 NC 1 1 4.53 ND 0 1 4.44 OH 1 1 4.52 OK 1 1 4.41 OR 0 0 4.52 PA 0 0 4.50 RI 0 0 4.50 SC 1 1 4.45 SD 0 1 4.49 TN 1 1 4.49 TX 1 1 4.54 UT 0 1 4.48 VT 1 0 4.46 VA 1 1 4.56 WA 0 0 4.57 WV 1 1 4.36 WI 0 0 4.51 WY 1 1 4.55 50 2004 Data VPRPOP UR BLACK MARRIED HSMORE EDREG DOM0003 AL -0.213 13.64 29.49 18.53 79.26 0 42.86 AK 0.138 15.69 3.37 16.83 79.99 0 20.00 AZ -0.315 12.59 3.36 12.65 77.46 0 13.33 AR -0.574 14.43 20.45 21.46 84.19 0 54.29 CA -0.205 12.93 6.38 11.13 81.21 0 25.00 CO -0.023 12.47 4.87 15.23 80.99 0 2.86 CT -0.364 12.71 12.36 7.91 87.37 1 16.67 DE -0.270 7.58 23.17 9.62 81.62 0 23.81 FL -0.151 10.28 19.64 11.37 79.86 0 30.00 GA -0.142 9.66 33.60 16.02 77.80 0 7.14 HI -0.863 9.76 1.81 13.71 90.16 0 60.00 IL -0.359 12.44 17.25 9.81 83.19 0 10.17 IN -0.359 11.35 9.23 14.40 80.40 0 28.00 IA 0.500 10.14 2.20 16.22 86.21 0 16.00 KS -0.375 11.34 4.45 16.48 85.90 0 50.00 KY 0.232 13.55 7.13 17.17 81.50 0 15.79 LA -0.113 13.26 35.79 11.65 75.67 0 23.08 ME 0.372 8.74 0.75 8.43 84.01 1 2.86 MD 0.005 10.5 30.69 7.78 85.62 0 40.43 MA -0.172 10.63 8.58 5.05 87.14 0 70.00 MI 0.187 13.39 14.93 9.70 82.65 0 15.79 MN 0.788 9.11 4.56 9.32 86.00 1 5.97 MS 0.074 17.13 43.07 14.78 76.87 0 15.38 MO 0.073 12.52 13.40 15.34 81.47 0 17.65 NE -0.225 8.77 7.27 17.74 77.32 0 23.81 NH 0.228 8.65 1.07 7.09 84.92 1 50.00 NJ 0.016 9.43 14.77 6.84 84.95 0 10.00 NM -0.331 11.68 4.11 14.06 77.50 0 14.29 NY -0.167 12.17 18.51 5.29 83.32 0 19.35 NC -0.301 12.21 24.28 16.46 81.65 0 12.00 OH 0.215 12.73 12.91 12.70 83.66 0 33.33 OK -0.269 10.37 5.91 21.25 79.61 0 16.67 OR 0.000 17.01 1.62 13.73 81.44 0 3.33 PA -0.292 12.6 11.76 8.36 85.31 0 16.00 RI -0.258 10.87 5.87 4.13 87.23 0 68.42 SC -0.253 12.86 32.82 14.23 81.09 0 10.87 SD -0.050 7.95 0.90 17.55 83.21 0 48.57 TN -0.481 11.76 18.60 16.68 82.06 0 9.09 TX -0.468 12.55 11.91 15.56 78.16 0 22.58 UT 0.028 10.53 0.74 28.06 85.82 0 11.63 51 VPRPOP UR BLACK MARRIED HSMORE EDREG DOM0003 VA -0.401 8.77 21.60 13.30 85.25 0 20.00 WA 0.175 12.66 3.21 14.53 81.41 0 2.04 WV -0.166 11.3 2.11 16.05 83.47 0 41.18 WI 0.531 9.34 6.50 9.32 85.64 1 9.09 WY -0.080 8 1.37 21.62 85.00 1 33.33 RSTRVOTE REDST LOGGSP AL 1 1 5.11 AK 0 1 4.46 AZ 1 1 5.26 AR 1 1 4.87 CA 0 0 6.15 CO 1 1 5.26 CT 0 0 5.22 DE 0 0 4.67 FL 1 1 5.74 GA 1 1 5.49 HI 1 0 4.65 IL 1 0 5.69 IN 1 1 5.32 IA 0 1 5.00 KS 1 1 4.95 KY 1 1 5.08 LA 0 1 5.14 ME 0 0 4.59 MD 0 0 5.31 MA 0 0 5.46 MI 0 0 5.53 MN 0 0 5.31 MS 1 1 4.83 MO 1 1 5.27 NE 1 1 4.95 NH 1 0 4.68 NJ 0 0 5.57 NM 1 1 4.76 NY 0 0 5.93 NC 1 1 5.47 OH 1 1 5.59 OK 1 1 4.99 OR 0 0 5.11 PA 0 0 5.62 52 RSTRVOTE REDST LOGGSP RI 0 0 4.58 SC 1 1 5.08 SD 0 1 4.42 TN 1 1 5.29 TX 1 1 5.91 UT 0 1 4.86 VA 1 1 5.47 WA 0 0 5.36 WV 1 1 4.64 WI 0 0 5.28 WY 1 1 4.29 53 APPENDIX II Table 6: OLS: Voter Participation of Voters Age 18-24 ?? 2000 2004 Coefficient T- ratio Coefficient T- ratio Constant -3.416 -2.32 -0.65 -0.38 UR 0.009 0.33 -0.002 -0.11 BLACK 0.005 0.9 -0.006 -1.21 MARRIED 0.002 0.1 -0.016 -1.08 HSMORE 0.037 2.09 0.012 0.62 RSTRHAT -0.003 -0.02 0.062 0.38 EDHAT -0.133 -0.57 0.148 0.64 DOM0003 -0.003 -1.23 -0.008 -2.69 R-squared 0.12 0.24 F-statistic 0.81 1.7 No. of Obs. 48 45 * denotes significance at 0.05 percent level ?? ** denotes significance at 0.10 percent level ?? 54 Table 7: OLS: Youth Voter Participation 2000 & 2004 Coefficient T-ratio Constant -4.035 -5.88 UR 0.029 1.93 BLACK 0.0002 0.08 MARRIED -0.007 -0.65 HSMORE 0.045 5.43 RSTRHAT 0.083 0.8 EDHAT -0.026 -0.16 DOM0003 -0.006 -3.54 R-squared 0.36 F-statistic 6.94 No. of Obs. 93 * denotes significance at 0.05 percent level ** denotes significance at 0.10 percent level